The inversion of data from complex 3-D resistivity and I.P.
surveys
M.H.Loke K. Frankcombe
Geotomo Software ExpolreGeo
Penang, Malaysia., Wangara, WA, Australia
drmhloke@yahoo.com kim@exploregeo.com.au
SUMMARY
The search and recovery for base and precious metals in
recent years has led to surveys in more challenging areas
over complex deposits and in extreme terrains. Such
deposits frequently have accessory minerals that can be
detected by induced polarization (I.P.) surveys. Due to
their complex shapes and host terrains, 3-D surveys and
inversion models are necessary to accurately resolve
them. However, in some cases, the survey lines are not
arranged rectilinearly. To accommodate an arbitrary
arrangement of the electrodes, a model discretisation that
is independent of the electrode positions is used. The
rugged terrain can be accurately modelled by the use of
the finite-element method where the surface of the mesh
matches the topography. Innovative arrays such as the
offset pole-dipole array have been used to rapidly survey
large areas at a lower cost compared to traditional dipole-
dipole arrays. Such arrays frequently have large
geometric factors that make it difficult to accurately
calculate the I.P. anomalies with the conventional linear
perturbation approach that uses the difference of two
resistivity calculations. The complex resistivity method,
where the I.P. component becomes the imaginary
component of the resistivity model, avoids this problem
as it effectively decouples the resistivity and I.P.
calculations. Furthermore, time-lapse 3-D surveys using
surface and borehole electrodes have been conducted to
monitor the flow of sodium cyanide solution directly
injected in steep-sided ore rock piles for secondary
recovery of gold. A 4-D resistivity inversion method is
used to map the flow of the solution during the injection
process.
Key words: 3-D, resistivity, I.P., inversion, time-lapse
INTRODUCTION
The high price of base and precious metals in recent years has
led to surveys in more challenging areas with complex
geology and extreme terrains. Many mineral deposits have
accessory minerals that can be detected by induced
polarization (I.P.) surveys, which measures both resistive and
capacitive aspects of the rock. Due to their complex shapes
and host terrains, 3-D surveys and inversion models are
necessary to accurately resolve them. The availability of
practical 3-D inversion software for microcomputers have also
made it possible to reinterpret data from old surveys for a
more accurate re-appraisal of prospects. While modern 3-D
field surveys attempt to arrange the electrodes in a rectilinear
rd
23 International Geophysical Conference and Exhibition, 11-14 August 2013 - Melbourne, Australia
D.F.Rucker
Hydrogeophysics, Inc
Tucson, Arizona, USA
druck8240@gmail.com
grid along parallel lines, old 2-D survey lines (particularly
when acquired in multiple campaigns) frequently do not have
a simple pattern.
The high price of gold has led to secondary recovery efforts by
injecting sodium cyanide solution into ore rock piles. In order
to optimise the secondary recovery process, it is necessary to
monitor the flow of the solution during the injection process.
The following section describes different numerical methods
that are used to produce more realistic models from 3-D
surveys. This is followed by examples from two field surveys.
METHODS
This section briefly describes a few numerical techniques used
to model the data from complex 3-D field surveys.
Smoothness-constrained least-squares inversion
The smoothness-constrained least-squares optimisation
method is frequently used for 2-D and 3-D inversion of
resistivity data (Loke et al., 2003). The subsurface model
consists of a large number of cells so that any resistivity
distribution can be accommodated. The equation that gives the
relationship between the model parameters and the measured
data is given below.
J Ti Rd J i i WT Rm W ri J Ti Rd g i i WT Rm Wr i1 (1)
The Jacobian matrix J contains the sensitivities of the
measurements with respect to the model parameters, is the
damping factor vector and g is the data misfit vector. ri-1 is the
model parameter vector (the logarithm of the model resistivity
values) for the previous iteration, while is ri is the change in
the model parameters. W incorporates the roughness filters in
the x, y and z directions. Rd and Rm are weighting matrices
used so that different elements of the data misfit and model
roughness vectors are given equal weights if the L1-norm
inversion method is used (Loke et al., 2003).
Topography
Many mineral deposits occur in areas with extreme terrains
that must be accurately modelled by the forward modelling
routine used to calculate the model response. As the
subsurface resistivity and I.P. can have arbitrary distributions,
the finite-difference and finite-element methods are commonly
used. Figure 1 shows the arrangement of mesh used by the two
methods to model surface topography. As the finite-difference
method normally uses a rectangular mesh, the surface
topography is modelled by small rectangular steps (Figure 1a).
The mesh cells above the ground are assigned a very high
1
The inversion of data from complex 3-D resistivity and I.P. surveys Loke, Frankcombe amd Rucker

resistivity value (eg. 10000 times the ground resistivity) to lines. I.P. surveys along 2-D lines have been carried out since
simulate the air layer. The finite-element method is more the 1950's. Interpretation of the data was mainly qualitative
suitable as the position of the surface nodes can be adjusted to due to the lack of practical inversion software. In some areas,
match the topography (Figure 1a) so that the topography is old data has been reinterpreted using modern software for a
directly incorporated into the inversion model (Loke, 2000). more accurate (and low cost) re-appraisal of old prospects.
The surveys were rarely carried out along parallel lines, and
frequently the lines have different directions. To interpret such
data, it is necessary to use a flexible model discretisation that
is not directly tied to the electrode positions (Figure 2). Each
model block is subdivided by four mesh lines in the x and y
directions (Figure 3). If the electrode falls on a node location
(at the intersection of the mesh lines), it can be directly
modelled by that node. There are two alternatives to model an
electrode at a position when it does not fall on a mesh node.
The first is by interpolating (Spitzer, 1999) the potentials
(Figure 3a) at the four nearest nodes in the mesh (and
Figure 1. Topography modelling using the (a) finite-
similarly replace a current electrode by four equivalent current
difference and (b) finite-element methods.
sources). The second method moves the nearest node to the
location of the electrode using a distorted finite-element mesh
Large I.P. effects
(Figure 3b). This method can be used if the distance between
two electrodes is more than the mesh spacing.
There are two methods used to calculate I.P. effects, the
perturbation (Oldenburg and Li, 1994) and the complex
resistivity (Kenma et al., 2000) methods. The first method
assumes the intrinsic I.P. values are sufficiently small so that a
linear perturbation of a base resistivity model can be used.
Consider a base model that has a conductivity DC. The effect
of the chargeability m is to decrease the effective conductivity
to IP = (1 - m) DC. The apparent I..P. (ma) is then calculated
by two forward models using the potentials () from the
original and perturbed conductivities.
ma = [ (IP ) - (DC ) ] / (DC) (2)
While perturbation approach works well in most cases, it has
two main problems. Firstly it is based on the assumption that
the intrinsic I.P. values are sufficiently 'small'. The second and
more serious problem is that it's accuracy depends on the
accuracy of two DC potentials. The difference is usually less
than 1% of the potential values, so it magnifies numerical
errors in the finite-difference or finite-element method used to
calculate the DC potentials. For many arrays, this is usually
not a problem. However, the offset pole-dipole and dipole-
dipole arrays (White et al. 2001) can have very large
geometric factors. In some situations, with large resistivity
contrasts, this can sometimes lead to negative apparent
resistivity values (Jung et al., 2009). In such cases, the
calculated apparent chargeability values are not reliable. The Figure 2. Example of surveys lines in different directions.
second I.P. model calculation method is to treat the
conductivity as a complex quantity with real and imaginary Time-lapse surveys
components (Kenma et al., 2000), which is given by
= DC i mDC . (3) In some surveys, repeated 3-D measurements are carried out to
The DC conductivity DC forms real part, while mDC forms detect temporal changes in the subsurface. The temporal
the imaginary part. A complex potential, with two components changes in the resistivity are frequently much smaller than the
r and i, is then calculated. spatial variations. Using the difference in models from
independent inversions frequently display artefacts due to
= r + i i (4)
noise. To reduce the artefacts, a 4-D inversion methodology
The apparent chargeability is calculated using the ratio of the (Loke et al., 2013) that directly incorporates the time domain
imaginary component to the real component, ma = i /r. The with the space domain is used. The least-squares equation is
accuracy of the apparent I.P. values does not depend on the modified to the following form.
accuracy of the D.C. potential. J T
i
R d J i i WT R m W M T R t M ri
(5)
Non-rectilinear survey grids
T
J Rdgi
i W R
i
T
m W iM
T
R Mr
t i 1

M is the difference matrix applied across the time models with

Modern 3-D surveys use an arrangement with the electrodes only the diagonal and one sub diagonal elements having
along a series of parallel lines. However, in areas with extreme values of 1 and -1, respectively. It minimises the difference in
topography, it might be necessary to deviate some of the the resistivity of each model cell and the corresponding cell

rd
23 International Geophysical Conference and Exhibition, 11-14 August 2013 - Melbourne, Australia 2
The inversion of data from complex 3-D resistivity and I.P. surveys Loke, Frankcombe amd Rucker

for the next temporal model. is the temporal damping factor
that gives the relative weight for minimising the temporal
changes in the resistivity compared to the model smoothness
and data misfit.
Figure 3. Methods to model the effect of an electrode not
on a node using (a) interpolation and (b) distorted grid
methods.
the surface topography are shown in Figure 6b. Due to the use
of a radial layout, the data coverage is very sparse towards the
edges of the model grid (Figure 6a). As resistivity distribution
within the ore heaps is highly inhomogeneous, the change in
the resistivity is used to monitor the flow of the solution.
Figures 6b and 6c shows the results from one series of
measurements in the form of iso-surface contours for the -4%
change in the resistivity at different times. Note the area with
the largest change is located to the north of the well. This is
probably due to differences in the subsurface permeability and
structural nonuniformities within the heap created during end-
dump construction (Rucker et al., 2013). The heap was built
up over the past 20 years by trucks dumping fresh ore over the
edge of older ore.

RESULTS

This section presents the results from two field surveys with
complex geology and unusual field arrangements.

Burra, South Australia

The Burra copper deposit was discovered in 1845 and mining

started in 1848 and ceased in 1877. It was at one time the
largest copper mine in Australia. It was reopened in 1971 and
closed again in 1981. Figure 2 shows some of the lines from a
1966 I.P. survey. Because of urban development and
consequent restricted access for any new survey, a re-
interpretation of the data was carried using modern 3-D
inversion methods to glean more information from it. There is
highly uneven data coverage. As such, the model use smaller
50x50 m. blocks in the northern part and larger 100x100 m.
blocks towards the sides in the southern part. The inversion
model shows a significant I.P. anomaly at the x-y location of
about (49900,101400) at depths of between 100 to 200 m.
(Figure 4b). At the same location, there is a region of
generally lower resistivity values of about 40 to 90 ohm.m.
This corresponds with the Eagle prospect, currently being
drilled by Phoenix Copper, nearly 50 years after the survey!
The north-south trending linear resistivity low, pfe high,
coincides with the Kingston Fault and defines the line of lode
in the Burra field. The nature of the high I.P. anomalies
towards the bottom-left edge of the deeper layers is uncertain Figure 4. Resistivity and I.P. models for the Burra survey
as there is not much data coverage there. They enticingly lie shown in the form of layers. The depths given are distances
on interpreted structures with the eastern most anomaly lying from the ground surface.
in the Kingston Fault, but neither have been tested. Figure 5
shows the resistivity and I.P. models as 3-D plots.
CONCLUSIONS
Cripple Creek, Colorado
The search for mineral resources in increasingly more
Sodium cyanide solution was injected into ore rock piles for challenging environments using 3-D I.P. surveys require
secondary recovery of gold after surface leaching had ceased parallel improvements in data interpretation techniques to
at the Cripple Creek and Victor Gold Mine in Colorado, USA provide more realistic subsurface models. The use of the
in September 2011. Resistivity measurements were made to smoothness-constrained least-squares inversion method, and
monitor the flow of the solution so as to optimise gold the finite-element method for areas with topography, enable
recovery. The measurements were made using 48 electrodes the use of sufficiently fine model discretisations that can
on the ground surface arranged in a radial pattern (Figure 6), match the complex geology. The use of the complex
94 electrodes along six boreholes and 8 long electrodes using resistivity method to calculate I.P. effects avoids the problem
steel-case injection wells (Rucker et al., 2013). Each snapshot of numerical errors in the forward modelling routine when
took 14 minutes to complete, and a total of 780 snapshots arrays with large geometric factors are used. A 4-D time-lapse
were acquired. The positions of the electrodes together with inversion methodology using smoothness constraints in both

rd
23 International Geophysical Conference and Exhibition, 11-14 August 2013 - Melbourne, Australia 3
The inversion of data from complex 3-D resistivity and I.P. surveys Loke, Frankcombe amd Rucker

the spatial and temporal domains enables the accurate
mapping of temporal changes.
ACKNOWLEDGMENTS
We would like to thank AngloGoldAshanti for the permission
to use Cripple Creek data set, and Phoenix Copper for the
Burra data set.
Amendment Experiment. Near Surface Geophysics (in press).
Spitzer K., Chouteau M. and Boulanger O. 1999, Grid-
independent electrode positioning for 3D DC and IP forward
modeling: Proc. 2nd. Internat. Sym. 3D Electromagnetics,
189192.
White, R.M.S., Collins, S., Denne, R., Hee, R., Brown, P.,
2001, A new survey design for 3D IP modelling at Copper
Hill: Exploration Geophysics 32, 152-155.

Figure 5. 3-D views of the Burra model showing (a)

resistivity as rendered volume, (b) I.P. as isosurfaces with
9, 12 and 15 % PFE.

REFERENCES

Jung, H.K., Min, D.J., Lee, H.S., Oh, S.H., and Chung, H.,
2009, Negative apparent resistivity in dipoledipole electrical
surveys: Exploration Geophysics 40, 3340.

Kenma, A., Binley, A., Ramirez, A. and Daily, W., 2000.

Complex resistivity tomography for environmental
applications. Chemical Engineering Journal, 77, 11-18.

Loke, M.H., 2000, Topographic modelling in resistivity

imaging inversion: 62nd EAGE Conference & Technical
Exhibition Extended Abstracts, D-2.

Loke, M.H., Acworth, I., Dahlin, T., 2003, A comparison of

smooth and blocky inversion methods in 2D electrical
Figure 6. Cripple Creek survey. (a) Overhead view of the
imaging surveys: Exploration Geophysics 34 , 182-187.
inversion model grid with electrodes layout. (b) Iso-surface
contours for the -4% resistivity change at different times
Loke, M.H., Dahlin, T. and Rucker, D.F., 2013, Smoothness-
after the injection of the sodium cyanide solution (that
constrained time-lapse inversion of data from 3-D resistivity
started at 2.8 hours from the first data set in snapshots
surveys. Near Surface Geophysics (in press).
used). t1= 1.1 hours, t2= 2.4 hours, t3= 3.7 hours, t4= 4.9
hours. (c) Overhead view of iso-surfaces.
Oldenburg, D. W., and Li, Y., 1994, Inversion of induced
polarization data: Geophysics 59, 1327-1341.

Rucker, D.F., Crook N., Winterton J., McNeill M., Baldyga

C.A., Noonan G. and Fink, J.B., 2013, Real-Time Electrical
Monitoring of Reagent Delivery during a Subsurface

rd
23 International Geophysical Conference and Exhibition, 11-14 August 2013 - Melbourne, Australia 4